Browser Bridge

web MCP Server

[Unmaintained] Browser automation MCP bridge for Claude Code. Rust.

Verified
webweb
5 views0 stars0 forksMIT

Why This Matters

Discovered via github-topic:mcp and last synced 4mo ago.

Verified
Source
github-topic:mcp
Stars
0
Last synced
4mo ago
Install
Check source

Install

Install instructions not detected yet

Check the source repository for the latest setup steps.

View source instructions
1
Tools
0
Resources
0
Prompts
Standard I/O
Transport

Available Tools (1)

_Linux_

_Windows-000.svg)](#) </div> > Open to contributors. The 21 MCP tools work today across Chrome / Brave / Edge / Arc. The rough edges are cross-platform testing and a workflow layer. See [Help Wanted](#help-wanted). ## Why this exists Headless browser automation breaks constantly on real-world sites: bot detection, login walls, fingerprint checks. Meanwhile your actual browser has none of those problems because it's logged in and looks human. I wanted Claude to be able to just drive that browser the way I would. Browser Bridge is a single Rust binary with two modes. Mode one (`serve`) runs as an MCP server over stdio so Claude Code can talk to it. Mode two (auto) runs as a native-messaging host that a Chrome extension launches. The two halves meet on a Unix socket (or localhost TCP on Windows), and every tool Claude calls gets translated into a Chrome DevTools Protocol command against your real browser. Result: Claude can read any page you're logged into, click things, fill forms, screenshot, watch network traffic, execute JS, record GIFs, pierce Shadow DOM. All without a separate browser session. ## Tools Claude gets 21 MCP tools (`src/mcp/tools.rs` + `src/mcp/tools_advanced.rs`): - **Navigation** - `navigate`, `tabs_context_mcp`, `tabs_create_mcp`, `switch_browser` - **Interaction** - `computer` (screenshots, clicks, typing, scroll, keyboard, drag, hover, zoom), `form_input`, `upload_image`, `resize_window` - **Reading** - `read_page` (accessibility tree with stable refs, Shadow DOM piercing, computed styles), `get_page_text`, `find` (natural-language element search) - **Scripting** - `javascript_tool`, `dom_query` (CSS selectors, pierces Shadow DOM) - **Observability** - `read_console_messages`, `read_network_requests` - **Storage** - `storage` (localStorage / sessionStorage / cookies get/set/delete/list) - **Synchronization** - `wait_for` (element visible/hidden, text match, network idle, JS predicate) - **Recording** - `gif_creator` (animated session exports with click overlays) - **Meta** - `update_plan`, `shortcuts_list`, `shortcuts_execute` (shortcuts framework is skeleton - see [Help Wanted](#help-wanted) #2) ## Architecture ``` Claude Code